: '"!''. <7 ';.';; ":.
. ATE SCHOOL
SECURITY CL ASSI f \C ATlON O^ THIS P A f. f r»T.»o Data Cnl.r*rtj
REPORT DOCUUEN'TATtGU PAGE
i re fo«T numbI K
:.::;•■/. -^ completing form
2. OOVT ACCSIS4ION NO
J. RECIPIENT - * C AT ALOG NUUtLH
4 TITLE fand Judllil.)
Automatic Welding Control Using a State
S TYfJ Of REPORT ft PERIOD COVERFO
0. PERFORMING ORG. HI^OOT NUMSCK
a. CONTRACT OR GRANT NUMBERf.j
MOODY, WILLIAM V.
I ped'OKuinj organization nauC anO ADDRESS
Massachusetts Institute of Technology
10. PROGRAM EL Em^nT, PROJECT TASK
ARZa 6 WORK UNIT NUMBERS
II CONTROLLING Of ICE NAME anO ADDRESS
NAVAL POSTGRADUATE SCHOOL
MONTEREY, CALIFORNIA 93940
12. REPORT DATE
II. NUMBER OF PAGES
MONITORING AGENCY NAME * AOORESSrif rttttorwnt Irom Controlling Olllct)
IS SECURITY CLASS, tot Inla report)
iTT DECLASSIFICATION/ DOWNGRADING
l« DISTRIBUTION STATEMENT tot thl, R»porl)
APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED
17. DISTRIBUTION STATEMENT tot <*- mbmtrmel mrtlorod In Blook 20, II dltlaronl from* Report)
l« SUPPLEMENTARY NOTES
l» KEY WORDS (Contlnuo on tmrmrmm aid* II nmco»*mrr and Identity by block ALMtoar;
Modern Control Theory
20 ABSTRACT (Continue on rovormo aid* It nacaaaary and Idontlty by block msmbmr)
DD I * 71 1473 EDITION OP 1 MOV «• IS OBSOLETE
tOnnrn, 1 \ S/N 1 1- 1 4 - t> A0 1
Although automatic welders have existed commercially for at least the past ten
years, they have proved inadequate to fulfill the needs of current shipbuilding,
marine structure and major pipeline manufacturing contractors. The recent
Alaskan pipeline construction program where hundreds of skilled manual welding
craftsmen were sought out and sent to our most northern state is a good example.
The development of a reliable and adaptable automatic welding process capable of
rapidly producing good weldments is considered to be a primary requirement of
Presented here is a description of the process variables encountered in the Gas
Tungsten Arc and the Gas Metal Arc Welding processes; the description emphasis
is placed on the variable interdependence which occurs in these processes. From
these variable relationships, a ninth order non-linear state variable description
of the Gas Metal Arc process is developed using nine first order non-linear
differential relations. Further definition of the exact nature of these rela-
tions will permit the development of a second generation automatic welder which
will be a dramatic improvement over existing machines.
This work is believed to be the first attempt to apply modern control theory to
DD 1» U73 UAJtLAS
/- / vl rviAo t\ -\ a ft r A i ' ' ' '■ L - -■■'-■'■ ■ ■
Approved for public release;
AUTOMATIC WELDING CONTROL USING A STATE VARIABLE MODEL
WILLIAM VINCENT MOODY
B.S. Ocean Engineering, U.S. Naval Academy
Submitted in Partial Fulfillment
of the Requirements for the
MASTER OF SCIENCE IN OCEAN ENGINEERING
and the Degree of
MASTER OF SCIENCE IN MECHANICAL ENGINEERING
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
(c) William Vincent Moody 197 9
DUDLEY KNOX LIDRARY
NAVAL POSTGRADUATE SCHOOL
MONTEREY, CALIF, 93940
AUTOMATIC WELDING CONTROL USING A STATE VARIABLE MODEL
WILLIAM VINCENT MOODY
Submitted to the Department of Ocean Engineering and the
Department of Mechanical Engineering on May 11, 1979, in
partial fulfillment of the requirements for the Degree of
Master of Science in Ocean Engineering and the Degree of
Master of Science in Mechanical Engineering.
Although automatic welders have existed commercially
for at least the past ten years, they have proved inadequate
to fulfill the needs of current shipbuilding, marine structure
and major pipeline manufacturing contractors. The recent
Alaskan pipeline construction program where hundreds of skilled
manual welding craftsmen were sought out and sent to our most
northern state is a good example. The development of a reliable
and adaptable automatic welding process capable of rapidly pro-
ducing good weldments is considered to be a primary requirement
of today's industry.
Presented here is a description of the process variables
encountered in the Gas Tungsten Arc and the Gas Metal Arc
Welding processes; the description emphasis is placed on the
variable interdependence which occurs in these processes.
From these variable relationships, a ninth order non-linear
state variable description of the Gas Metal Arc process is
developed using nine first order non-linear differential
relations. Further definition of the exact nature of these
relations will permit the development of a second generation
automatic welder which will be a dramatic improvement over
This work is believed to be the first attempt to apply
modern control theory to welding.
Supervisor: Koichi Masubuchi
Title: Professor of Ocean Engineering and Materials
Supervisor: Henry Martyn Paynter
Title: Professor of Mechanical Engineering
I would like to express my appreciation to Professor
Koichi Masubuchi and Professor Henry Paynter for the instruc-
tion, contributions and suggestions that I received both in
the courses which they taught and in their support of this
work. Without their guidance, this thesis would not have
achieved its final form.
The real purpose of scientific method is to
make sure Nature hasn't misled you into thinking
you know something you don't actually know. . . .
If you get careless or go romanticizing scientific
information, giving it a flourish here and there.
Nature will soon make a complete fool out of you.
— Robert M. Pirsig
Zen and the Art of Motorcycle Maintenance
TABLE OF CONTENTS
Title Page 1
Table of Contents 5
List of Figures 6
List of Tables 8
I. INTRODUCTION 9
II. WELDING PROCESS VARIABLES 14
II. A. Control Variables 20
II. A. 1. Arc Voltage and Current 2
II. A. 2. Shielding Gas 27
II. A. 3. Pulsing Shape and Frequency .... 28
II. A. 4. Traverse Speed 31
II. A. 5. Electrode Tracking Path 32
1 1. A. 6. Welding Torch Movement 33
II. A. 7. Wire Feed Rate 35
II. A. 8. External Magnetic Fields
Surrounding the Arc 3 5
II. B. State Variables 36
II.B.l. Puddle Shape and Size (Penetration) 36
II. 3. 2. Metal Temperature Distribution. . . 36
II. B. 3. Arc Temperature, Arc Shape
and Composition 37
II. B. 4. Arc Length 38
II. B. 5. Metal Transfer Mode (and Rate) ... 38
II. B. 6. Arc Magnetic Field 43
II. B. 7. Radiant Light Emission 49
II. B. 8. Acoustic Emissions 49
II. C. Summary 53
III. STATE VARIABLE MODEL 55
III. A. Simplifying Assumptions 55
III.B. Model Variables 57
III.C. Model Formulation 58
IV. RECOMMENDATIONS AND CONCLUSIONS 64
V. COMMENTS REGARDING AUTOMATIC WELDING 67
LIST OF FIGURES
Figure Title Page
2-1 Block Diagram for Mechanized Gas Metal
Arc Welding Systems 16
2-2 Block Diagram for Mechanized Gas Tungsten
Arc Welding Systems 17
2-3 Drooping Voltage Characteristics in Welding
Power Supplies 21
2-4 Constant-Voltage and Increasing-Voltage
Characteristics in Welding Power Supplies 23
2-5 Constant-Current Characteristics in Welding
Power Supplies 24
2-5A Voltage Characteristics of a Welding Arc 25
2-6 Gas Shield Configuration Developed at David
W. Taylor Naval Research and Development Center 2 9
2-7 Current Pulsing 31
2-8 Possible Horizontal Electrode Tracking Path
Positioning System 34
2-9 Globular Transfer Mode 4
2-10 Axial Spray Transfer Mode 41
2-11 Short Circuit (or Dip) Transfer Mode 42
2-12 Transition Current in GMA Welding 45
2-13 Shift in Transition Current Magnitude with
Material 4 6
2-14 Shift in Transition Current Magnitude with
Electrode Extension (Stick-out)
2-15 Shift in Transition Current Magnitude with
2-16 Noise Emission Characteristics in Manual
Stick Electrode Welding 51
2-17 Possible Acoustic Emission Spectral Density 52
3-1 State Variable Model for an Automatic GMA
Welding Process 63
LIST OF TABLES
Table Title Page
2-1 Control and State Variables in the GMA
and GTA Welding Processes 15
2-2 Shielding Gas Applications 27
2-3 Fundamental Mechanisms of the GMA Metal
The label "Automatic" has been attached to many
different welding processes which have been developed by
industry or discussed in technical literature over the last
ten years; each of these automatic processes has attempted
to reduce the decision making or manipulative functions
played by the human welding operator by increasing the
"intelligence" of the machine. However, the success of the
above mentioned processes is dubious if the goal of auto-
mation is to develop a truely versatile, adaptable and
dependable welding process. The automatic machines currently
in operation have narrow and extremely specific ranges of
Two broad categories of automatic welders are employed
in present industrial applications: "robot" welders and
modified manual welders. Seemingly, each new issue of the
popular welding journals contains at least one description
of a new automatic machine which falls into one of the above
categories. Automation systems of various stages of com-
plexity have been designed and applied to welding processes
such as Gas Metal Arc (GMA) , Gas Tungsten Arc (GTA) , Resis-
tance, Electroslag and Submerged Arc. However, with the
exception of certain assembly line type production processes
such as the automobile industry, manual production welding
completely dominates industry worldwide. One only has to look
at current shipyard or pipeline production operations to
verify this fact.
Successful automation of welding processes is attrac-
tive for one or more of the following reasons:
•Cheaper production costs
•Reduced on-site manpower requirements
•Reduced welder training costs
•Increased production rates
Weld Quality Improvement
•Provide near ideal welding conditions
•Weldment consistency and reproductiveness
•Immediate post-welding non-destructive testing
•Lower number of weld repairs
To realize these substantial advantages over manual welding,
a second generation automatic machine must be developed.
Current automation systems employed on welders consist
of some combination of open loop and closed loop control
schemes. An open loop control scheme can be as simple as
presetting the value of one process variable. The machine
then attempts to match this preset value without any measured
feedback. More advanced machines measure one or more of the
variables in the welding process and, through the use of a
controller, attempt to minimize variations of these variables
from preset values. To illustrate the difference between an
open loop and a closed loop control scheme, the variable
weld head (electrode) position is a good example. In certain
pipeline manufacturing systems, the welding machine consists
of multiple GMA welding heads which ride on a chain or steel
band-type track that is attached circumf erentially around
the pipe. This is an open loop form of control because the
track is preset and invariant; the success of this scheme
depends on how carefully the operators prepare the joint and
allign the track. Closed loop control of the weld head
( electrode ) position has been implemented on GMA machines
designed to butt weld two flat plates. In this case, sensors
such as a TV camera or a mechanical probe placed ahead of the
electrode sense the track of the joint path and position the
electrode via servomotor drives. By using feedback in this
manner, joint irregularities can be overcome by the welding
machine because the electrode tracking system can follow an
imperfect joint matchup.
Current automatic welders are successful only under
specific applications. The major problem with these current
machines is that only selected process variables are monitored
or controlled. Because there are a relatively large number of
variables in any welding process, the relevance to the final
weldment quality of the variables selected for control is
critical. For the most part, the controlled variables in
current machines include arc length, arc voltage or current,
current pulsing, wire feed rate, electrode tracking path,
and electrode traverse speed. Therefore, the successful
machine designer has been forced to choose the right combina-
tion of these variables to control in order to produce a
Recent advances in control theory are predicated on
the ability of the control engineer to successfully model
a process so that all pertinent process variables can be
identified and measured (or at least estimated) ; then the
process variables are controlled by minimizing any variation
about a desired set point. To date modern control theory has
not been applied to welding processes for a variety of rea-
sons; the primary hurdle thus far has been the inability to
develop a successful model.
This work is the preliminary step of a multi-year
project jointly headed by Professor Koichi Masubuchi,
Professor Henry M. Paynter, and Mr. Frans Van Dyck at MIT
to develop a fully automated, versatile and reliable welding
process. This work will first identify a set of process
variables taken from the Gas Metal Arc and Gas Tungsten Arc
Welding processes. These processes were chosen because they
appear to provide the broadest range of possible applications
Secondly, each variable will be discussed thoroughly with the
emphasis placed on the variable's suitability in a feedback
control system. Finally, problems such as measurement,
estimation, and processing of each variable will be
The ultimate goal of this work is to provide the
basis for a model of the GMA (or GTA) process and then to be'
gin to "close the loop" in the control formulation. Once
this first step is completed, the project will continue to
develop a variable controlled welding process. This com-
pleted welding process, when coupled with immediate non-
destructive weldment testing (NDT) and post-testing repair,
will provide a rapid and reliable method for fabrication of
such structures as U.S. Navy and commercial ships, pipeline
systems, and large ocean platforms while incorporating the
advantages listed earlier in this introduction.
II. WELDING PROCESS VARIABLES
Unlike many systems and processes that are controlled
today, welding presents a major problem because the desired
output is not measurable. The goal of any welding process is
to produce a weldment that matches the base metal in chemical,
metallurgical and mechanical properties. However, weldment
qualities such as porosity, amount of fusion, hardness, tough-
ness, tensil strength, etc., cannot be determined "on-line"
while the welding operation is in progress. Only after the
process is completed can a weldment sample be taken and tested
for its qualities. Therefore, other properties of the welding
system must be monitored and from these properties, projections
and estimations of the weldment qualities can be made. These
weldment quality projections are based upon empirical rela-
tionships that have been obtained from the millions of
experiments completed by welding researchers.
Using this empirical data base, weldment quality can be
controlled by measuring or estimating and then controlling a
set of process variables. In the following discussion, this
set of process variables will be broken up into two subsets : the
control (or manipulative) variables and the state variables. The
control (manipulative) variables are defined as those variables
that can be manipulated directly by the welding machine or the weld-
ing operator. The state variables are defined as those variables
which are determined by the welding process itself once the
control (manipulative) variables have been set. The state variables
cannot be manipulated directly but are functions of the welding process.
Figures 2-1 and 2-2 are block diagrams of the Gas Metal
Arc and Gas Tungsten Arc welding processes. These diagrams
identify the principal components of these systems. Through
the use of these figures, the reader can gain an appreciation
for the complexity of the welding processes and can identify
how the following listing of state variables and control (manipulative)
variables are related schematically. This list (Table 2-1)
is meant to be as complete as possible and was developed from
a similar list contained in Appendix C of reference 20.
CONTROL AND STATE VARIABLES IN THE GMA AND GTA WELDING PROCESSES
Control (manipulative) Variables
Voltage (AC or DC and
Pulsing Frequency and Shape
Electrode Tracking Path
Welding Torch Movement
Wire Feed Rate (GMA)
Filler Wire Feed Rate (GTA)
Composition and Flow Pattern
of Protecting Gas
External Magnetic Fields
Puddle Shape and Size
Metal Temperature Distribution
Arc Shape and Composition
Metal Transfer Mode
Arc Magnetic Field
Radiant Light Emission
Electrode Extension Length (GMA)
BLOCK DIAGRAM FOR MECHANIZED
GAS METAL ARC WELDING SYSTEMS
I—— — ■— | W^JM. ' .** ILUiU
BLOCK DIAGRAM FOR MECHANIZED
GAS TUNGSTEN ARC WELDING SYSTEMS
>*»■ i «i ■■-»■»*«■
To define a control scheme for these processes, first
a brief discussion of nomenclature is required. Initially a
simplified diagram will be proposed which will later be ex-
panded into a final state variable model. The control vector
U will be an m-component vector containing part or all of the
above listed control variables. The state vector X will be
defined as an n-component vector containing part or all of
the above listed state variables. The final choice of the
components of the control and state variable vectors should
be the minimum set of variables that accurately describe all
of the important functions of the welding process.
These vectors can be mathematically represented as
U = vector of m control
X = vector of n state variables =
L n J
2- at 5
Then the relationship between X and U can be diagrammed as
where F(X, U, t) is a set of time dependent, non-linear
equations describing the welding process. By adding a
controller to this representation, a closed loop process
X = F(X, U, t)
For the remainder of Section II, each control and
state variable will be discussed in detail.
II. A. 1. Arc Voltage and Current
The magnitude of voltage and current supplied by
the power supply in the welding process is directly related
to the amount of heat input that is supplied to the weldment
once power losses are taken into consideration. Power
supplies currently used in the GTA process have a "drooping"
characteristic which supply a maximum open circuit voltage
and then decrease to a zero value of short circuit voltage
(Figure 2-3). Power supplies used in GMA processes have
constant or increasing voltage characteristics (Figure 2-4) .
FIGURE 2-3: Drooping Voltage Characteristics in Welding
L = arc length
L 2 > L x
P = open circuit voltage
T = short circuit voltage
(Masubuchi, 18, p. 2-46)
The constant voltage power supply allows the GMA
process to self-regulate arc length because weld current,
arc voltage and arc length are interrelated in the following
1. Arc length vairation in the GMA process is
minimized if the melting rate (or burn-off rate)
of the consummable electrode (inches per second)
is equal to the feed rate of the electrode.
2. Arc length is directly proportional to arc voltage —
increasing arc length increases arc voltage.
3. Melting rate increases with current increase.
For example, if the wire-feed speed increases, the arc
length and arc voltage decreases, causing an increase in
current. 1 This increase in current increases the melting
rate and burns-off the electrode more quickly; therefore the
arc length increases (Masubuchi, 18, p. 2-49). Schaper, in
reference 30, has discovered experimentally that welding
performance in the GMA process is further improved if the
power supply is capable of providing both the characteristics
of constant voltage and constant current (Figure 2-5) by
operating over the entire volt-ampere range; an example of
such a machine is the TEK-TRAN linear slope control power
supply type LSC 750.
For the majority of applications in both the GMA
and GTA processes, direct current is used. Alternating
current has limited, specific applications such as the
FIGURE 2-4: Constant-Voltage and Increasing Voltage
Characteristics in Welding Power Supplies
a. CONSTANT-VOLTAGE CHARACTERISTIC b. INCREASING "VOLTAGE CHARACTERISTIC
L = arc length
L 2 > L x
P = open circuit voltage
Q = unstable arc operating point
S = stable arc operating point
(Masubuchi, 18, p. 2-48)
FIGURE 2-5: Constant Current Characteristics in Welding
L = arc length
L 2 > L
p = open circuit voltage
T = short circuit voltage constant current value
(Masubuchi, 18, p. 2-46)
FIGURE 2-5A: Voltage Characteristics of a Welding Arc
D.C. POWER SUPPLY RESISTANCE 2!
CATHODct. . •:.;>'.. •
V A (ANOOE VOLTAGE DROP)
VplARC COLUMN VOLTAGE DROP)
V c (CATHODE VOLTAGE DROP)
(Masubuchi, 18, p. 2-6!
application of high frequency in the GTA process to improve
the arc stability. In direct current applications, two
choices of polarity are available and are labeled Direct
Current Straight Polarity (DCSP) and Direct Current Reverse
Polarity (DCRP) .
Direct Current Straight
Direct Current Reverse
DCSP is used in the majority of GTA welding applications
because approximately 80% of the heat is developed at the
Anode in a welding arc and 5% of the heat is developed at
the Cathode. Primarily, the use of DCRP in GTA welding
occurs when Aluminum or Magnesium are welded; the oxide
coating on these metals is removed much more readily when
using DCRP by an unknown mechanism (possibly the positive
ions striking the plate) (Masubuchi, 18) .
Conversely, DCRP is the most common polarity used in
GMA welding. DCRP provides a faster melting rate of the
electrode and, consequently, a higher metal transfer rate.
II. A. 2. Shielding Gas
The primary consideration regarding the shielding
gas is the correct choice of the right gas mixture; also,
these gas mixtures must be free of impurities because any
gas impurities can cause arc instabilities. These gas mix-
tures have been determined experimentally for different
applications and are summarized in the table below:
■ ■ i
Base Metal Gas
Steel + 1% 0„
Carbon C0~ or
Titanium Argon DCSP
co 2 +o 2 ,
o 2 , co 2 ,
or C0 2 +0 2
Quenched & Argon + DCRP
The purpose of the shielding gas is to provide a
shield from atmospheric contamination which causes porosity,
embrittlement, cracking, and other problems. Schaper
(reference 30) has developed a gas shield configuration
which has significantly increased the ability of the
shielding gasses to exclude the atmosphere from the weldment.
This new shield has proved capable of dispersing the gas
thoroughly and is diagrammed in Figure 2-6.
Potential dynamic uses of the shielding gas have been
discussed but not confirmed experimentally. For example,
it may be possible to control the temperature on the outside
arc boundary or even to position the arc by directing the
flow of the gas. However, these dynamic uses may be invalid
for outdoor applications on windy days even if they are
II. A. 3. Pulsing Shape and Frequency
Current pulsing has applications in both GTA
(Troyer, 33) and GMA (Lesnewich, 17) welding processes; the
effects of pulsing for welding control are substantial and to
date have not been completely documented. For example, in
the GTA process pulsing can be used advantageously to agitate
the weldment puddle to improve fusion and prevent cracking.
However, the desirability of pulsed current is best understood
by examining its effect in GMA applications.
FIGURE 2-6: Gas Shield Configuration Developed at David W.
Taylor Naval Research and Development Center
1/16" copper sum
■SMtfio and surreal positions
J.I.11UU1UJ.U UJ.U-.U U_JJ.U1_IJ.UJ_.
- WAT I RUNE
/A" COPPFR TUBF
if! \])\ j Ffli : fi j if! lp -^^^
ft " I ^K'^'-:
-jj.jui iliMiliiJ :ji;ili!)ji!!;i ijn r rrnrti m n /,.'
ttli r I nil- h n<r .•ssaji.HS!;-^
GAS I INE -VENTED
SECTION A A
1/16" COPPFR SHEET
LL-Lr . . j.--i
SECTION B B
(Schaper, 30, p. 6)
The metal transfer mode in GMA is determined by the
magnitude of the applied current. To achieve spray transfer
(metal transfer modes will be discussed later) , the magnitude
of the current must reach a certain level. However, once
this transition current value is reached, the metal transfer
rate and the depth of penetration increase tremendously and
are difficult to control. Therefore, to maintain an average
low applied power and to control the metal transfer rate,
current pulsing is used. Pulsed-power supplies provide a
low average current but at the peak of the square wave pulse
exceed the transition current (Figure 2-7) . Consequently,
the metal transfer rate and depth of penetration can be
controlled by choosing the frequency and magnitude of the
Although pulsing is a promising variable with which
to control both the GMA and GTA welding processes, much more
developmental work must be completed. Currently, the magni-
tude and shape (square wave) of the pulses are pre-programmed
into the machine. However, once a larger empirical data base
is obtained, active frequency and pulse shape variations may
be implemented as part of a closed loop control scheme.
-S^N+- _ _ —
(Lesnewich, 17, p. 4)
II-. A. 4 Traverse Speed
Electrode traverse speed is inversely proportional
to the amount of heat energy applied to the weldment. As the
traverse speed decreases, the heat input increases. Most current
machines implement a constant traverse speed. However, by
measuring variables such as weldment puddle width and metal
temperature distribution, traverse speed can be controlled
and become a useful variable.
II. A. 5. Electrode Tracking Path
To reduce joint preparation costs and to mini-
mize joint allignment problems, some form of closed-loop
tracking system should be incorporated as was discussed
in the introduction of this study. Evans (reference 7)
has performed a fairly extensive survey of electrode
tracking systems in use through 1974. Generally, these
systems are equipped with either mechanical or optical
sensors which follow the joint. For example, the commerci-
ally available "Arc-Tender" described by Evans uses a mech-
anical probe that is inserted into the joint ahead of the
torch and "feels" the path much like a blind person might
use a cane.
Optical photocell sensors have also been incorporated
which also have the added advantage of sensing and controlling
arc length (Evans, 7, p. 18). However, the two most promising
methods of sensing the joint path appear to be either a
television (or video scan) camera or the Cecil mechanical
cross-slide tracking and sensoring system.
The Cecil system has been used successfully in ship-
yards and has most recently been used by Schaper for a
narrow gap welder that is being developed at the David W.
Taylor Research and Development Center. The system consists
of a transducer controlled finger probe which is placed
about 1/4 inch ahead of the torch and senses irregularities
in the joint sidewall. It can operate in joint thicknesses
from to 2 inches (Schaper, 30, p. 6) .
The television tracking system is attractive because
it would not only provide joint path information but could
also provide information regarding the weldment puddle width.
In case of either the mechanical, transducer or television
sensor, joint irregularities would be used to position the
electrode via some form of servomotor drive. Probably the
most useful configuration for the tracking system would in-
volve mounting the torch on a carriage drive that follows a
preset track; this track could be horizontally or vertically
attached in the case of plate welding or could be circum-
ferentially attached in the case of pipeline or cylindrical
welding (Figure 2-8) . The servomotor positioning drive could
then have the capability of positioning the torch ±2 inches
(or as much as required) from the centerline of the preset
II. A. 6. Welding Torch Movement
The tracking system described above would give
the torch two degrees of freedom. Additional degrees of
freedom added to the torch would allow it to more closely
duplicate the motions used in manual welding. Using the
terms of Naval Architecture, the tracking system provides
control of surge (electrode traverse speed) and sway
FIGURE 2-8: POSSIBLE HORIZONTAL ELECTRODE PATH POSITIONING
Electric Power emdOGas Feed
(servomotor positioning) . Additional controls of heave
(vertical) , roll (angular rotation about the axis parallel
to the joint path) , and pitch (angular rotation about the
axis parallel to the sway axis) may be useful.
II. A. 7. Wire Feed Rate
In most GMA applications, wire feed rate is preset and
invariant. However, wire feed rate could be made variable
and would be useful for controlling arc length. Experiments
are needed in this area.
II. A. 8. External Magnetic Fields Surrounding the Arc
Jose Converti has suggested a form of magnetic
arc control that he has begun to test experimentally using
the GTA process (Converti, 13). By positioning 4 magnetic
sensors at 90° intervals circumferentially about the arc,
arc position and rate of change of arc position can be
detected. Then, through the use of a set of deflection coils,
the arc can be positioned in the joint. Possible other
advantages of this scheme may be the control of arc shape
and the control of arc instabilities. Dynamic control of
the arc in this manner may have some useful consequences.
II. B. State Variables
II.B.l. Puddle Shape and Size (Penetration)
Vroman and Brandt (reference 34) have shown that
the weldment puddle width can be measured using a video scan
camera and this measurement can be used to control the torch
traverse speed. The puddle width was preset into the controller
and the welding process attempted to maintain this width
throughout the length of the weldment. Good results were
obtained from this experiment with the weldments displaying
a high degree of uniformity.
Vroman and Brandt suggested that this puddle width
information also be used to manipulate the welding current.
Limitations on the use of this variable are based primarily
upon the accuracy and resolution capability of the video scan
camera. However, puddle width is a good indication of the
amount of heat input to the weldment and the metal deposition
II. B. 2. Metal Temperature Distribution
By employing a temperature sensor such as a
thermocouple at the front and back side of the weldment,
indications about depth of penetration and heat input can
be obtained. Bennett (reference 3) and Smith (reference 32)
have indicated that this is a viable means of control. Some
practical implementation considerations exist however;
for example, access restrictions to the back of the weldment
are encourtered in pipeline welding or in the repair of
large, high pressure steam lines. However, a NASA develop-
ment in 1967 of a "hybird" thermocouple system for GTA
aluminum welding proved that useful information could be
obtained from the torch side of the joint only (Evans, 7,
p. 18). The hybird thermocouples consisting of a constantan
wire which contacted the aluminum near the arc. Holding the
voltage constant, either the torch travel speed or the current
were varied to maintain a constant temperature. The following
conclusions were made from this NASA project:
•Thermocouple insensitive to radiated arc
interference and variations in surface emissivity.
•Surface oxides produced no interference.
•More successful for aluminum plate thicknesses
of 1/8 inch or less.
•No observable lag in response time.
II. B. 3. Arc Temperature, Arc Shape and Composition
Arc temperature can also be detected by some form
of micro- thermistor or micro-pyrometer . However, many factors
affect the arc and much more study regarding arc physics is
required. For example, the principal arc temperature is
determined by the choice of the type of shielding gas. The
ionization potential for Helium (24.5 eV) is higher than the
ionization potential for Argon (15.7 eV) ; therefore the
Helium arc temperature is significantly higher than the
Argon arc temperature. The arc dynamics probably hold much
useful information for the welding engineer but the physical
mechanism is not fully understood at this time.
II. B. 4. Arc Length
Arc length is a critical variable in the welding
process and must be controlled precisely. As stated previously,
arc length is proportional to arc voltage and may also be
regulated by the wire feed rate. Other information regarding
arc length may be contained in the arc temperature emissions,
light emissions and acoustic emissions ;, the acoustic emission
relationship will be discussed below.
II. B. 5. Metal Transfer Mode (and Rate)
The metal transfer mode in the GMA process can
take one of three forms: globular, spray (axial or rotating),
or short-circuit (dip) . The transfer mode is a complicated
phenomenum and is a function of many factors such as:
•Shielding gas type
•Polarity (DCRP, DCSP)
•Electrode diameter (d)
•Electrode extension or stick-out (L)
•Electrode metal type (steel, Al, etc.)
•Activation of the electrode with alkali, alkaline
earth and rare earth materials (Masubuchi, 18)
The Globular mode (Figure 2-9) involves the transfer
of large volume drops of metal (typically larger in diameter
than the electrode) from the electrode to the weldment at
the rate of 3 or 4 per second. These drops travel at a
relatively low velocity. The Axial spray mode is the most
desirable for most applications and consists of many small
volume metal droplets that are accelerated to high velocities
in a direction parallel to the length of the electrode. This
mode deposits metal directly into weldment; it is the most
efficient and stable mode. When the welding current exceeds
a certain value, the electrode tip begins to rotate and
sprays the metal at extremely high velocities in a cornical-
type pattern. This rotating spray mode is highly unstable
and should be avoided.
For certain applications, short-circuit or dip transfer
is desirable. For example, dip transfer can be used to
increase the depth of penetration when welding steel. In dip
transfer, a large metal globe is formed on the end of the
electrode which simultaneously contacts the weldment; this
globe then separates from the electrode and is accelerated
into the weldment (Figure 2-11) .
FIGURE 2-9: Globular Transfer Mode
FIGURE 2-10: Axial Spray Transfer Mode
FIGURE 2-11: Short Circuit (or Dip) Transfer Mode
,. Change of Electrode Position
b. Current v? Tin*
c. Voltage vi Tim«
(Masubuchi, 18, p. 3-29)
The amelioration of the transfer mode from globular
to axial spray occurs at a value of current known as the
transition current. At this value of current, the drop
volume decreases significantly and the drop velocity increases
(Figure 2-12) . The value of current required in the transition
region varies with the electrode material, the electrode
diameter and the electrode extension. The electrode extension
length determines how much resistance heating occurs as the
electric power is conducted through the electrode; as the
length increases, the resistance increases and the resistance
heating increases. As the conductivity of the material
increases, extension length becomes less of a factor. From
a qualitative viewpoint, the transition current magnitude
decreases as extension length increases and increases as
electrode diameter increases (see Figures 2-13, 2-14, and 2-15).
The metal transfer mode is a very critical variable
in GMA welding and should be one of the primary considerations
in any control scheme. Table 2-3 summarizes the important
parameters affecting this variable.
II. B. 6. Arc Magnetic Field
As discussed previously in the control variable
section, the magnetic characteristics of the arc can be
sensed and maybe used to deflect the arc. Other information
may be obtained from this variable but extensive experiments
must be completed first.
FUNDAMENTAL MECHANISMS OF THE GMA METAL TRANSFER MODE
TYPE OF TRANSFER MATERIAL POLARITY
FIGURE 2-12: Transition Current in GMA Welding
1 „-^ l"
■ ie o
200 MO «C0
(Masubuchi, 18, p. 3-17)
FIGURE 2-13: Shift in Transition Current Magnitude with
FIGURE 2-14: Shift in Transition Current Magnitude with
Electrode Extension (Stick-out)
£ = electrode extension
*1 > £ 2
FIGURE 2-15: Shift in Transition Current Magnitude with
d 2 > dl
II. B. 7. Radiant Light Emission
The light emission spectrum is one of the few
variables listed in this section that is used as feedback
by the manual welder. The brightness and intensity of the
light is of sufficient magnitude that rapid damage occurs in
the welder's eye unless protective shielding is worn. How-
ever, the ability of the welder to process anything other
than "brightness" and arc length information from the light
emissions is debatable. With the use of more sophisticated
equipment, spectral information may be recoverable from the
light. Experiments have been performed in this area and a
patent titled "Control System Using Radiant-Energy Detection
Scanning" (US Patent Number 3,370,151 of 20 February 1968)
has been issued to Neil J. Normando (Vroman, 34, p. 749).
Glickstein (reference 9) also lists two references pertaining
to this topic. From the results of further experimental work,
light emissions could prove to be a useful variable.
II. B. 8. Acoustic Emissions
When observing a welding craftsman at work, one is
amazed by what appears to be a lack of sensory feedback.
Anyone who has attempted to initiate a shielded metal elec-
trode arc while wearing the protective shield can appreciate
the lack of visual feedback. However, the experienced
welder still produces good welds.
Apparently, one of the feedback mechanisms that the
welder uses is the crackling sound of a "good" weld. Welding
shop foremen are able to monitor the progress of their pro-
duction welders by listening to the sound; the sound has been
described as the same sound bacon makes when frying in a pan.
In a short series of experiments conducted with Mr.
Anthony J. Zona at the MIT Welding Lab, a correlation between
the noise emissions and the arc length was observed. By
assuming the noise emissions to be band limited white noise,
the following figure (Figure 2-16) approximates the nature of
the acoustic output during welding. These characteristics
are based upon observations made during manual stick electrode
welding of steel plate.
From these observations, the following conclusions
•The central, or dominant, frequency (f ) shifted
to a lower value as the arc length increased.
•As the arc length decreased, the frequency
shifted to a higher value.
•Acoustic energy decreased as the arc length
increased and increased slightly as the arc
Although the acoustic energy is probably dependent
upon many factors, metal transfer rate appears to be an
Noise Emission Characteristics in Manual Stick Electrode
w N (f)
Actual spectral characteristics of the acoustic
spectrum would have a more general shape; hopefully, the
spectral density curve will have the appearance of Figure 2-17
Second order affects may be present in the form of peaks in
the curve, but the curve as shown in Figure 2-17 has a
dominant peak at f and then decays with frequency.
Possible Acoustic Emission Spectral Density
w N (f)
By attaching a directional transducer to the torch,
these noise emissions could be measured and sent to a micro-
processor. If the mean square value of the noise is the pri-
mary important parameter in these emissions, the signal could
be processed quite easily. However, if the frequency distri-
bution of the noise is important (observations indicate it is
important) , spectral density processing techniques would be
The instruments described in reference 1 have the
capability of performing real time spectral density analysis
over a band width of about five octaves. They have a sampling
frequency of about 60,000 Hz and cost about $30,000 (1978 US)
which may be fairly excessive for welding applications. How-
ever, if the noise is narrow banded about the peak frequency
f , perhaps a cheaper instrument with a narrower band width
could be fabricated.
If Figure 2-17 is the general appearance of the noise
spectrum, optimal values of f would need to be obtained
experimentally for each welding application. For example,
according to Mr. Zona, the noise characteristics seem to
change when welding in wet weather. This is a very compli-
. cated area and will require a lot of developmental work.
II. C. Summary
The variable descriptions in this section are designed
to give the reader an appreciation for the complexity of the
welding process; particular emphasis has been placed on the
interactions that occur between the variables. Only through
an understanding of this variable interaction can the control
engineer begin to develop a model and a control scheme for
welding. The references contained at the end of this work
provide a more complete description of the fundamentals of
welding; in particular, reference 18 discusses in-depth many
of the variables that have been presented here.
Although an effort has been made to identify every
potential welding variable that is important to an automatic
welder, current knowledge limitations restrict the use of
some of them in a model of the process. In the following
section/ an automatic welder model will be developed which
encorporates ten of the variables listed in this section.
III. STATE VARIABLE MODEL
In this section, a state variable model will be
developed to describe the GMA welding process. The GMA
process is considered to have more applications in major
ocean structural production and is therefore more important;
however, a similar model could be envisioned for the GTA
III. A. Simplifying Assumptions
Because the total number of variables described in
Section II is enormous, certain simplifying assumptions must
be made to produce a meaningful model. These assumptions are
based on one or both of the following:
•Lack of sufficient data or knowledge regarding
the variable and its affect on the overall
•The procedure required to control the variable
is well defined currently and this variable can
be added to the model easily.
For the above reasons the following variables have been
neglected from consideration in the final model:
•Electrode Tracking Path
•Welding Torch Movement
•External Magnetic Fields
•Arc Shape and Composition
•Arc Magnetic Field
•Radiant Light Emission
The electrode tracking path can be controlled using
existing technology and has no affect on the model once the
assumption is made that the electrode remains in the joint.
The major problem that will be encountered in controlling the
movement along the joint path by the electrode is the choice
of a suitable sensor that produces accurate and rapid position
Likewise, welding torch movement and manipulation can
be implemented once the movement requirements have been
defined. For the following model, the torch will have one
degree of freedom — forward transversing or surge. Sway
(sideways) motion is assumed but only in conjunction with
the electrode remaining in the joint path.
For the remainder of the excluded variables listed
above, not enough knowledge is currently available to determine
the affect of implementing them in the control scheme. As
further research is completed, these variables may be added
to the model .
Finally, the model assumes that the type and diameter
of the electrode have been chosen, that the shielding gas
(Argon) has been chosen and that the material being welded
has been chosen. Also, Direct Current Reverse Polarity is
III.B. Model Variables
The remaining variables are listed below along with
a mathematical symbol which will be used in the model
development. These variables are:
Control (or Manipulative) Variables
•Traverse speed v
•Pulsing frequency co
•Wire feed rate C
•Puddle (weldment) width w
•Metal temperature T
•Arc length L_
• Electrode Extension Length L
These variables are assumed to be deterministic and
measurable. Power supply characteristics are assumed to be
III.C. Model Formulation
Taking the variables as defined in Section IILB. , the
following vectors are identified.
U = Control (or Manipulative) Variable Vector =
X = State Variable Vector
R = Reference vector (initial set points
of the control vector)
! "i 1
Y = Output vector =
L A I
! L T
The output vector contains the four components of the
state vector plus m which is the metal transfer rate; m is
m = C,,(I,u> /L_) (3.1)
5 p E
Now by taking two equations proposed by Professor
Masubuchi in his 13.17J course at MIT, the definition of the
nature of the state relations can begin. These equations
are as follows:
V = A 1 L A + T X L A + C l (3 ' 2)
(Masubuchi, 18, p. 2-31)
M_ = A„I + B L„I 2 (3.3)
r 2 2 E'
(Masubuchi, 18, p. 3-52)
A,, B, , C, , Ay and 3- are constants. The symbol d is the
electrode diameter and the symbol M is the melting rate.
Then these relationships may be established:
ft (L E ) = ^E = F 4 ( meltin g rate, to, C) (3.4)
j£-(L fl ) = L & = F 3 (melting rate, a , C, V) (3.5)
ir(V) = V = G 2 (L A , L A , I, I, V ) (3.6)
Using the variable interaction descriptions contained
in Section II, the remainder of the non-linear differential
relations may be defined as follows:
|_(I) = i = G 1 (a) , I, V, L E , m, I q ) (3.7)
f^v) = v = G 3 (I, L E , u , w, T M , v q ) (3.8)
ar ( V = ^p = G 4 (I ' V V w ' m ' V ) (3,9)
lt (C) = b = S (L A' V ' V *' C } (3 - 10)
_( W ) = w = p (Melting rate, u> , v) (3.11)
ft ( V = *M = F 2 (V ' lf V V) (3 * 12)
By substituting L„ and I into the above equations
for the melting rate and substituting for the derivative
terms in the right hand side of the above relations, the
final model equations can be listed below.
I = G 1 (o) , I, V, L £/ m, I Q ) (3.13)
V = G 2 [F 3 (I,L E ,aj p ,C,V) ,1^,1^(0) , I, V,L E ,m) ,V q ] (3.14)
v = G 3 (I, L E , U) , w, T M/ v q ) (3.15)
oo = G„(I, L_, oj , w, m, oj ) (3.16)
p 4 E p po
C = G 5 [L A ,V,P 3 (I,L E/ aj ,C f V) ,G 2 [F 3 (I,L E ,a3 ,C,V),
L^I/G-^ai ,I,V,L E ,m) ] ,C q ] (3.17)
w = F. (L , I, a) , v) (3.18)
T M = F 2 (V/ I ' V V) (3.19)
L A = F 3 (L E / X ' V C ' V) (3.20)
L E = F 4 (L E' Z ' %' C) (3.21)
The output vector becomes:
Y l = C l ( ^ = w (3.22)
Y 2 = c 2 (X) = T M (3.23)
Y 3 " C 3 (X) = L A (3.24)
Y 4 = C 4 (X) = L E (3.25)
Y 5 = 05(1, uj L E ) = m (3.26)
The above relations define the ninth order model and
the output. These relationships are diagramed finally in
IV. RECOMMENDATIONS AND CONCLUSIONS
The model developed in the previous section is the
first step in the implementation of an optimal control
scheme for the GMA welding process. The following steps
must be taken prior to the development, design and construc-
tion of an actual automatic welder using this model
•The nature of the vector functions F(X,U) ,
G(Y/U/R) and C(X,U) must be determined.
•Power supply characteristics must be known
exactly. Preferably, a power supply capable of
operating anywhere on the V-I plane such as the
one described by Schaper (reference 30) will be
•Sensors accurate enough to monitor the process
variables must be developed.
•Values for the proper settings of the parameters
in R must be determined for each application.
•Correct choices of parameters such as electrode
diameter, electrode composition, joint alignment
tolerance and shielding gas composition must be
specified for each application.
The nature of some of the functional relations has
been addressed in the literature but for the most part,
only a few of the variables have been considered in the
equation formulation. Little is known how such variables
as pulsing frequency, metal temperature and weldment puddle
width affect the exact form of the equation formulation.
Secondly, even if the form of the equations does not vary
with changes of metal being welding, electrode diameter,
electrode composition and shielding gas, the constants and
values of any exponents probably do. The first derivation
for these equations should be accomplished considering steel
plate welding in Argon gas shielding because this situation
is the most prevalent for industrial needs.
The power supply characteristics will specify the
operating regions available for the welding process; the
power supply will be required to provide the current,
voltage and pulsing characteristics as requested by the
control scheme and is therefore the most important
component in the process. Every effort should be made
to make the power supply as adaptable as possible.
The settings of R have been researched extensively
and are fairly well defined. R will vary with the choice
of all parameters and will determine the operating set
points of the process.
Likewise, parameter requirements have been fairly
well defined in the literature for many GMA applications.
These requirements are based on extensive empirical data.
Existing sensors have been discussed and, with the
recent developments in electronic and video technology,
should be available for welding control. These sensors
should be able to provide sufficient accuracy.
Automatic control of the GMA process will entail
a very complicated implementation procedure. Presented
here has been a description of all identifiable process
variables and a mathematical method of bookkeeping to define
the variable interraction using modern control theory
In the following section, comments will be made
regarding the implementation of this control scheme to
welding. These comments are based on the authors ' s
observations, research and experience and in some cases
have not been documented by existing scientific evidence
V. COMMENTS REGARDING AUTOMATIC WELDING
The most important variable in the GMA welding process
when using DCRP Polarity appears to be the current; both
current magnitude and the frequency of the pulsing determine
the majority of the physical mechanisms which occur. Current
should be maintained so that the transition current is ex-
ceeded and axial spray transfer is maintained. Current and
pulsing determine the melting rate of the electrode and
affect the amount of heat energy that is applied to the
Therefore, the time rate of change of the current
should be determined by measurements (and hence be a
function) of most of the variables in the process. To do
this, the power supply has to be adaptable enough to operate
in the desired region. Schaper (reference 30) indicated that
the quality of the weldments that he produced improved
greatly once he shifted power supplies from a constant-
voltage type to one that provided constant-current
If the current can be maintained in the region that
insures axial spray transfer, the arc stability will increase
and the mechanical properties of the weldment should be of
good quality and should be reproducable.
Initial implementation of variable control for auto-
matic welding will probably be completed on a model that is
much simpler than the one given in Section III or on a model
of the GTA process. The GTA process would be simpler to
model because wire feed rate, melting rate and metal transfer
rates can be ignored.
If the GMA process is chosen, one or two of the state
variables (weldment puddle width and metal temperature) that
are not easily measured can be deleted and then model
development can proceed. Once computer model simulations
have been completed, testing on actual welding machines
would be advisable. After this first step is completed,
other variables can be added to the model. Eventually one
or two stochastic variables (light and noise emissions)
should be added because they probably contain much useful
information. These stochastic variables should unlock much
of the mystery of the dynamic characteristic of the arc
itself. Likewise, magnetic sensing and control of the arc
will furnish more information about the arc mechanism.
Two different types of automatic welders are envisioned
for the future. One type would be a large scale production
machine that is controlled automatically in the joint path
and is manipulated with little or no human intervention.
This machine would probably be very large, very heavy, and
very expensive. However, another type of machine can be
envisioned that could overcome the man-machine interface
problem and use the manual skills as an integral part of
the control scheme. This machine would be much smaller
(maybe even portable) and less costly and would have uses
such as repairs and on-site production.
Because manual manipulative skills and human sensor
(eyes, ears, touch) capabilities are difficult to duplicate
by machines, it seems tragic to eliminate them from welding.
A human operator could ensure that the torch followed the
joint path. The human operator could vary the traverse
speed; even such a simple system as three "idiot" lights
meaning "speed up," "slow down," and "OK" may be sufficient
to control traverse speed. The machine could then mani-
pulate other variables in the process. Welder training costs
and training time would be reduced substantially while
weldment quality would increase. As Houldcroft indicates
in his article, the welder could be compared to the pilot
in a commercial airline; the machine functions as an
extension of the operator to greatly improve the overall
State variable controlled automatic welding has
enormous potential to improve the fabrication capabilities
of industry worldwide. As structural designs have become
more complex and intricate, designers have demanded more
efficient joining methods. The development of a variable
controlled automatic welder should help to fulfill this need.
1. B & K Instruments, Inc., Technical Review to Advance
Techniques in Acoustical, Electrical and Mechanical
Measurement. Vol. 1, 1978, Denmark.
2. Barkow, A.G., "Recent Advances in the Field of Automatic
Welding of Pipeline Girth Welds in the USA, "
Proceedings of the Pipe Welding Conference 10-13
November 1969 , Paper 13. The Welding Institute,
Abington, England, 1970.
3. Bennett, A. P., "The Interaction of Material Variability
upon Process Requirements in Automatic Control,"
Advances in Welding Processes , Third International
Conference, 7-9 May 1974, Paper 3. The Welding
Institute, Abington, England, 1974.
4. Bryson, A.E. , Jr. and Y. Ho, Applied Optimal Control .
New York, 1975.
5. Danhier, F.G., "Automatic Orbital Welding on Pipework
and Pipelines," Proceedings of the Pipe Welding
Conference 10-13 November 1969 , Paper 3. The
Welding Institute, Abington, England, 1970.
6. Downey, J.C., D.H. Hood and D.D. Keiser, "Shrinkage in
Mechanized Welded 16-inch Stainless Pipe,"
Welding Journal , Vol. 54, No. 3, (March 1975),
7. Evans, R.M. , Automation in Welding . Metals and Ceramics
Information Center, Report No. MC1C-74-24.
Battelle, Columbus, Ohio. October, 1974.
8. Gavrilov, A.N. , Automation and Mechanization of Production
Processes in the Instrument Industry , trans. L.
Herdon. Oxford, England, 1967.
9. Glickstein, S.S. and W. Yeniscavich, "A Review of Minor
Element Effects on the Welding Arc and Weld
Penetration," Welding Research Council Bulletin ,
No. 226, May, 1977.
10. Hedrick, J.K., "2.152 Modern Control Theory and Applica-
tions; Course Notes," MIT, Spring 1977.
11. Houldcroft, Peter T. , "1977 Comfort A. Adams Lecture:
Developing Precision Assembly by Welding,"
Welding Journal , Vol. 56, No. 8, (August 1977),
12. Hosier, William, "Arc-Welding Times Slashed by Robot at
Arc Research," Welding Journal , Vol. 58, No. 1,
(January 1979), pp. 26-29.
13. Interview with Jose Converti, Ph.D. Candidate, Mechanical
Engineering Department, MIT, 7 December 1978.
14. Jones, R.L., D.K. Kiltau and K.P. Havik, "Automatic
Welding of Marine Pipelines on the Semac I Lay
Barge," Welding Journal , Vol. 57, No. 9,
(September 1978), pp. 15-21.
15. Karnopp, D. and R. Rosenberg, System Dynamics: A Unified
Approach, New York, 1975.
16. Kwakernaak, H. and R. Sivan, Linear Optimal Control
Systems, New York, 1972.
17. Lesnewich, A., "MIG Welding with Pulsed Power," Welding
Research Council Bulletin, No. 170, February, 1972
18. Masubuchi, K. , "13.17J Welding Engineering: Course Notes,"
MIT, Fall, 1978. (unpublished)
19. Masubuchi, K. , H.M. Paynter, and F. Van Dyck, "Preliminary
Research Proposal on 'Integrated Pipeline Manufactur-
ing System, ' " Laboratory for Manufacturing and
Productivity, MIT, Cambridge, Mass. (Unpublished
20. Masubuchi, K. , H.M. Paynter, and F. Van Dyck, "Preliminary
Research Proposal on ' Integrated Pipeline Manufactur-
ing System,' Second draft," Laboratory for
Manufacturing and Productivity, MIT, Cambridge,
Mass. (Unfinished Draft)
21. Mills, G.S., "Fundamental Mechanisms of Penetration in
GTA Welding," Welding Journal , Vol. 58, No. 1,
(January 1979) , pp. 21-s - 24-s.
22. Morris, G.J. and R.L. Apps, "Programmed Control of MIG
Welding Parameters," Welding and Metal Fabrication ,
Vol. 45, No. 8, (October 1977), pp. 487-492.
23. Morris, G.J. and J. Lowery, "How Electronically Controlled
MIG can cut Costs," Welding and Metal Fabrication ,
Vol. 46, No. 10, (December 1978) , pp. 687-6§9.
24. Murray, D.M., Applications of Aerospace Technology in
Industry. A Technology Transfer Profile: Welding.
Technology Management Group, ABT Associates, Inc. ,
Cambridge, Mass. September 1971.
25. "Narrow-Gap Welding Nears the Shipyards," Welding Design
and Fabrication, (March 1979) , pp. 118-120.
26. Ogata, K. , Modern Control Engineering , Englewood Cliffs,
New Jersey, 1970.
27. Paton, E.O., "Welding in the USSR: Technology and
Production," Welding and Metal Fabrication ,
Vol. 46, No. 3, (April 1978), pp. 201-207.
28. Phillips, Arthur L. , Current Welding Processes , ed.
A.L. Phillips. New York, 1964.
29. Quigley, M.B.C., "Physics of the Welding Arc," Welding and
Metal Fabrication , Vol. 45, No. 10, (December 1977) ,
30. Schaper, V.D., Development of an Automated Narrow Gap
Welder for High Strength Steel Plate , David W.
Taylor Research and Development Center, Report No.
MAT-76-86. Bethesda, Maryland, August 1977.
31. Schultz, D.G. and J.L. Melsa, State Functions and Linear
Control Systems, New York, 1967.
32. Smith, C.J. , "Self-Adaptive Control of Penetration in a
Tungsten Inert Gas Weld, " Advances in Welding
Processes , Third International Conference, 7-9
May 1974, Paper 41. The Welding Institute,
Abington, England, 1974.
33. Troyer, Wade E., "Programming and Pulsing the Tungsten
Arc," Metal Progress , Vol. 197, No. 3, (March
1975) , pp. 99-106.
34. Vroman, A.R. and H. Brandt, "Feedback Control of GTA
Welding Using Puddle Width Measurement, " Welding
Journal , Vol. 55, No. 9, (September 1976), pp.
35. Wilson, A.J. and D.T. Northcote, "The Development of
Semi- Automatic Welding of Pipe in Australia,"
Proceedings of the Pipe Welding Conference
10-13 November 1969 , Paper 34. The Welding
Institute, Abington, England, 1970.
c# l Automatic welding
control using a state
29 KA3 69
1 fEB 90
3 5 2 3
3 5 2 30
c'.l Automatic welding
control using a state
Automotive welding control using a state
3 2768 002 04737 5
DUDLEY KNOX LIBRARY
■■ ■ ■